NANO LETTERS
Selective Growth of Well-Aligned Semiconducting Single-Walled Carbon Nanotubes
2009 Vol. 9, No. 2 800-805
Lei Ding,† Alexander Tselev,† Jinyong Wang,‡ Dongning Yuan,† Haibin Chu,†,‡ Thomas P. McNicholas,† Yan Li,*,‡ and Jie Liu*,† Department of Chemistry, Duke UniVersity, Durham, North Carolina 27708, and Beijing National Laboratory for Molecular Sciences, Key Laboratory for the Physics and Chemistry of NanodeVices, National Laboratory of Rare Earth Material Chemistry and Application, College of Chemistry and Molecular Engineering, Peking UniVersity, Beijing 100871, China Received November 18, 2008; Revised Manuscript Received December 17, 2008
ABSTRACT High-density arrays of perfectly aligned single-walled carbon nanotubes (SWNTs) consisting almost exclusively of semiconducting nanotubes were grown on ST-cut single crystal quartz substrates. Raman spectroscopy together with electrical measurements of field effect transistors (FETs) fabricated from the as-grown samples showed that over 95% of the nanotubes in the arrays are semiconducting. The mechanism of selective growth was explored. It is proposed that introducing methanol in the growth process, combined with the interaction between the SWNTs and the quartz lattice, leads to the selective growth of aligned semiconducting nanotubes. Such a high density of horizontally aligned semiconducting SWNTs can be readily used in high current nanoFETs and sensors. This method demonstrates great promise to solve one of the most difficult problems which limits application of carbon nanotubes in nanoelectronicssthe coexistence of metallic and semiconducting nanotubes in samples produced by most, if not all, growth methods.
Single-walled carbon nanotubes have shown outstanding potential in nanoelectronics due to their exceptional properties such as high mobility,1 high thermal conductivity,2 and good chemical3 and mechanical stability.4 The device performance was shown to be better than that of Si-based devices in some applications.5 However, even though the devices made from individual nanotubes have shown outstanding performance,1,2,6 the high hope for the next generation of carbon nanotube based electronics is hampered by several major problems. Among them are the lack of reliable methods to control the alignment and position of nanotubes as well as, and perhaps most importantly, simultaneous growth of nanotubes with different structures (chiralities), yielding random mixtures of metallic and semiconducting nanotubes.7 Over the past decade, significant efforts in research have been focused on finding solutions to these problems but only limited progress has been made. For example, researchers have demonstrated the separation of metallic from semiconducting SWNTs based on electrophoresis,8 physicochemical modification,9,10 density gradient induced centrifugation,11 selective elimination by electrical breakdown,12 gas-phase plasma etching,13 and several other * Corresponding authors,
[email protected] and
[email protected]. † Duke University. ‡ Peking University. 10.1021/nl803496s CCC: $40.75 Published on Web 01/20/2009
2009 American Chemical Society
methods.14,15However, the chemical separation methods used in these processes always introduce defects and contamination to the SWNTs. Additionally, the separated nanotubes still need to be assembled onto a substrate for device fabrication. Currently no reliable methods exist to accomplish this step. Therefore the direct growth of metallic and semiconducting SWNTs on suitable substrates would be a significant advancement toward their easy integration with existing Si fabrication technology.16-22 According to these published papers, there is still some room to improve such as the ratio and alignment of the selectively grown SWNTs on surfaces. Nearly all the SWNTs of the selective samples were shown as random, which was thought to increase the overlap of the nanotubes and the resistance caused by the tube junctions. But the direct growth method was thought to be a significant approach for its advantages, such as high purity of nanotubes and the availability of methods to control the alignment of the nanotube during growth by electric field,23 laminar gas flow,24 and single crystal substrates.25-27 Recently, uniform nanotube arrays fabricated on single crystal substrates28,29 can be used directly as a thin film for making a large amount of devicessa major breakthrough in practical fabrication of nanotube devices. However, the coexistence of metallic and semiconducting
nanotubes in such samples is still a problem hindering their immediate application. Herein, we present a chemical vapor deposition (CVD) approach, which allows selective growth of high density arrays of well-aligned SWNTs with almost exclusively semiconducting SWNTs. The dense arrays of aligned SWNTs were grown on single-crystal ST-cut quartz substrates using an ethanol/methanol mixture as the carbon source and Cu nanoparticles as catalysts. Nanotubes within the arrays revealed a narrow diameter distribution from 1.55 to 1.78 nm with over 95% of nanotubes being semiconducting. We proved that the selective growth is a result of two crucial factors: the presence of methanol in the CVD process and a strong interaction (affinity) between SWNTs and the quartz lattice. The samples can be used for direct fabrication of nanotube FETs with good performances. Still, more work needs to be done to fully understand the growth mechanism and achieve reliable 100% selective growth of semiconducting nanotubes. However, the fact that such highly selective growth of well-aligned semiconducting nanotubes can be achieved represents a promise for solving the last major problem that is limiting the wide use of nanotubes in nanoelectronics. To the best of our knowledge, this is the first time that the control of horizontal alignment and the control of electronic properties can be achieved simultaneously. In our experiments, a 1.0 mM CuCl2/poly(vinylpyrrolidone) alcohol solution was used to deposit catalyst on to the substrate. The catalyst patterning and growth processes were similar to the method reported previously except for the addition of a methanol bubbler.29 Typically, ST-cut (36° Y-cut) single crystal quartz was used as the substrate for the SWNT growth unless otherwise specified. Other substrates including SiOx, Z-cut single crystal quartz, and 42° Y-cut single crystal quartz were also used in this study to help elucidate the growth mechanism. The growth experiments were performed in a 1 in. tube furnace at 900 °C. The substrate with catalyst precursor deposited on it was treated with oxygen plasma for 15 min to remove the poly(vinylpyrrolidone) in the catalyst precursor. Then the substrate was heated up to 800 °C and kept for 15 min with a flow of hydrogen (500 sccm), followed by CVD growth of SWNTs at 900 °C. Although the types of nanotubes grown under different ethanol/methanol ratios are not the same, typically, a flow of hydrogen (450 sccm) and argon (150 sccm through an ethanol bubbler and 300 sccm through a methanol bubbler) was used for the growth of semiconducting SWNTs. After 15 min of growth, the sample was cooled to room temperature and inspected with scanning electron microscopy (SEM), atomic force microscopy (AFM), and a Raman spectrometer. A mixture of metallic and semiconducting SWNTs can be obtained at different growth conditions on the same type of substrate. Resonant Raman spectroscopy (RRS) has been used to characterize the samples. RRS has been proven to be a powerful tool for characterizing and revealing the detailed structure and the electronic and phonon properties of SWNTs. The radial breathing mode (RBM) and G-band features of Nano Lett., Vol. 9, No. 2, 2009
Figure 1. Arrays of almost exclusively semiconducting SWNTs. (A and B) SEM images. The bright and parallel horizontal lines visible in the images are catalyst lines. (C) AFM image. (D) Diameter distribution of 200 SWNTs of an array measured by AFM. (E and F) Raman spectra of SWNTs transferred onto the SiOx/Si substrates. The spectra were obtained using 488 and 633 nm excitation laser lines at 10 different spots over the substrate for each laser line. Each curve in a panel shows a spectrum at a spot on the substrate. Peaks within the rectangles marked with S correspond to the semiconducting SWNTs. The rectangles marked with M denote the frequency range where RBM peaks of metallic SWNTs are expected.
Raman spectra can be used to assign and distinguish the semiconducting and metallic SWNTs, especially when multiple excitation lasers are used. In our experiment, lasers with three different wavelengths of 488, 633, and 785 nm were used to distinguish the semiconducting and metallic SWNTs sufficiently for our samples. More than 10 spots on each sample were characterized by RRS. In order to avoid the influence of the Raman peaks from the quartz substrates at 127 and 205 cm-1, the SWNTs grown on ST-cut quartz were transferred onto the SiOx/Si for RRS characterization (see Figure S1 in Supporting Information). Figure 1 presents the scanning electron microscopy (SEM) and atomic force microscopy (AFM) images and Raman spectra of the aligned semiconducting SWNTs. The uniform, high-density and perfectly aligned arrays of long nanotubes can be found on the entire ST-cut quartz substrate from SEM images. AFM measurements show that the nanotubes in the array have a narrow diameter distribution from 1.4 to 1.8 nm with an average diameter of 1.65 nm (Figure 1, panels C and D). The substrate surface is remarkably clean and free of amorphous carbon contamination. A high fraction of the semiconducting tubes in the arrays becomes evident from RRS data. Raman spectra recorded with 633 and 488 nm excitation lines (Figure 1, panels E and F), show a narrow 801
Figure 2. High on/off ratio FETs fabricated with as-grown aligned CNT arrays top-gated by solid electrolyte polymer films. (A) Large magnification view of the set of electrodes of a test device. The widths of the gaps between the 40 nm thick gold contact lines correspond to the channel lengths of FETs (from top to bottom): 4, 3, 2, 1, and 5 µm. To break possible current paths through the nanotube array past the transistor channels, the tubes were partially removed by reactive ion etching (RIE) in oxygen plasma. (B) Family of transfer characteristics (drain current, Id, vs top gate voltage, Vg) of one of the test devices under a bias voltage Vds ) 80 mV. The parameter of the family is the transistor channel length L. The channel width is 0.5 mm. (Inset) The on/off ratios obtained for the device as a function of the channel length, L.
distribution of RBM frequencies between 143 and 158 cm-1. Notably, no peaks are visible in the range, where RBM peaks of metallic tubes are expected from the so-called Kataura plot30 as highlighted in the figures with red rectangles. From the experimental relationship between a SWNT diameter d (in nm) and a RBM shift ω (in cm-1), ω ) 248/d,30 we estimated that the diameters of SWNTs range from 1.55 to 1.78 nm, which is narrower than that measured by AFM. It is possible that the deformation of SWNTs due to the force produced by the tip during AFM measurements leads to the apparent broadening of the diameter distribution. No RBM peaks have been found when a 785 nm laser was used for excitation, which further supports a low percentage of metallic SWNTs since no Raman peaks from semiconducting nanotubes are expected in the case of the 785 nm excitation laser because the diameter of the semiconducting nanotubes with resonance at this excitation line needs to be smaller (
